Signal Analyzer

• Single, dual, and multi-signal analysis modes
• Configurable evaluation targets (return, win rate, volatility)
• Save signals to Library for reuse in Strategy Builder
• Threshold calibration with visual breakpoints for isolating signal regimes
• Rolling and expanding window evaluation

• Use Signal Analyzer when the open question is whether an indicator has a stable relationship with a later outcome.
• Use Portfolio Backtest when you already know the rule and want to test it inside a portfolio.
• Use Strategy Builder when the next step is a signal-driven switching rule.

• Indicator: the predictor you test, such as price, SMA, EMA, RSI, trailing return, volatility, or drawdown.
• Signal asset: the asset used to compute each indicator.
• Transform: optional subtract or divide operation against a second indicator.
• Target ticker: the asset whose future behavior the forward metric measures.
• Signal return basis: whether indicators and target outcomes use total-return series or price-only raw closes. Total return is the default. Uploaded U.* aliases and unsupported .SIM raw series use total_return with a warning.
• Forward metric: the future outcome, such as forward return, win rate, or volatility.
• Horizon: the number of trading days between the signal observation and the measured outcome.

• Sampling frequency: how often the analysis samples the daily history.
• Signal lag: shifts the sampling schedule so you can check the same analysis against nearby alignments.
• Splits: divides the sample into time slices for stability checks.
• Fit method and bins: single-mode controls for polynomial fit or regressogram summaries.
• Metric selection: dual mode lets you switch the displayed heatmap metric while keeping the same sample.
• Tree controls: multi-mode settings such as train/test split, depth, minimum leaf size, and validation folds.

• Single mode: start with scatter and fit, then check diagnostics for horizon, sampling assumptions, correlation, and split coverage before reading bin means.
• Dual mode: use populated-cell coverage and best/worst cell diagnostics to separate broad effects from sparse outliers before saving a cell to the Library.
• Multi mode: compare train/test and cross-validation diagnostics, then inspect leaf-level means and rule paths before creating a signal from a leaf.

Signal Analyzer isolates a relationship so you can check it before embedding it in a strategy. It does not build the trading rule for you.

Save a run to retain reproducible inputs and diagnostics. Use the mode-specific selection control: single setup, dual cell, or multi leaf. Turn a validated pattern into a Library signal. Strategies and Strategy Builder can then use it.